Συντάχθηκε 01-02-2010 12:31
από Thekla Papadaki
Email συντάκτη: tpapadaki<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: σύνταξη/αποχώρηση υπάλληλος ΜΠΔ.
ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Τμήμα Μηχανικών Παραγωγής και Διοίκησης
Παρουσίαση Διδακτορικής Διατριβής
Τσαφαράκη Στέλιου
Δευτέρα 8 Φεβρουαρίου 2010, ώρα 11:00 π.μ.,
Αίθουσα συνεδριάσεων τμήματος Μ.Π.Δ.
Εξεταστική επιτροπή:
Ματσατσίνης Νικόλαος, Καθηγητής Τμήματος Μ.Π.Δ., Π.Κ. (Επιβλέπων Καθηγητής)
Τσουκιάς Αλέξανδρος, Διευθυντής Έρευνας CNRS, Université Paris Dauphine (μέλος τριμελούς)
Γρηγορούδης Ευάγγελος, Επίκουρος Καθηγητής Τμήματος Μ.Π.Δ., Π.Κ. (μέλος τριμελούς)
Πάσχος Ευάγγελος, Καθηγητής Université Paris Dauphine
Βλαχοπούλου Μάρω, Καθηγήτρια Πανεπιστημίου Μακεδονίας
Μυγδαλάς Αθανάσιος, Καθηγητής Αριστοτελείου Πανεπιστημίου
Μάνθου Βασιλική, Καθηγήτρια Πανεπιστημίου Μακεδονίας
Περίληψη
Designing optimal products is one of the most critical activities for a firm to stay competitive. The Optimal Product Line Design constitutes a wide area of research in quantitative marketing for over thirty years, which is usually formulated in the context of Conjoint Analysis. Dealing with this NP-hard combinatorial optimization problem, a manager must decide on a number of issues: how to simulate the consumer choice process, which optimization algorithm to apply, and how to model the possible retaliatory actions from competitors. The application of an effective approach has several important practical implications for marketing managers, since a bad designed product line may result in a lower than expected market share, or may even cannibalize the firm’s existing products. The manager should carefully compare the different alternative choice models, optimization algorithms, game theoretic approaches and choose those that better fit the company’s requirements. This constitutes a quite complex task, especially for marketing managers who usually do not have special knowledge concerning optimization algorithms and game theory. In this context, a number of marketing systems have been developed, assisting a manager in this problem of high complexity.
The marketing systems that have been developed so far use simple deterministic choice models for simulating the human choice process in order to reduce the problem’s complexity. The limitations of deterministic choice rules regarding the effectiveness of customer choice behavior simulation are well documented in the literature.
Furthermore, all marketing systems aim at improving the performance of a single best solution, which is finally provided to the decision maker. Product design however is a complex and not well formalized discipline that draws upon both marketing and engineering fields. In this context, while it is important for a firm to obtain the optimal solution, this product line configuration may not be technologically feasible, or the production cost may be prohibitive. Hence, it is just as critical for the managers to be provided with a wide range of near-optimal product profiles, in order to assess them using a number of secondary criteria such as production costs, strategic fit, and technological considerations.
Moreover, all the approaches that have been applied to the Optimal Product Line Design problem assume a static market, where the incumbent firms will not respond to the introduction of one or more new products. However, considering only competitors’ current products constitutes a very restrictive assumption, and the product designs derived from such static optimization approaches might proved to be optimal only for the short term. It is now becoming well known that in the longer run, optimization algorithms should take into account the retaliatory actions of the incumbent firms, which may launch new products or redesign their existing ones, as a response to the entrance of a new firm to the market.
In the present thesis, an integrated approach for dealing with the Optimal Product Line Design problem is presented, which combines state of the art methods for the three properties of the problem. The simulation of customer choice behavior is implemented through an innovative market simulation model that individually calibrates probabilistic choice rules. Particle Swarm Optimization, a new population-based algorithm inspired from natural intelligence, is used to provide a set of good near-optimal solutions, from which the decision maker will be able to select the most beneficial one. The concept of Nash equilibrium is employed for modeling the dynamic nature of competition, where each incumbent firm responds to competitive moves by redesigning its product line. All the methods and algorithms proposed are evaluated against the current state of the art approaches, using both simulated and real data regarding consumer preferences. The real data set was obtained from a market survey, the purpose of which was the measurement of customer preferences concerning milk. The proposed approach is the first attempt to integrate the three most important properties of the problem into a single methodology. A marketing system is developed, which integrates the underlying algorithms of the methodology under an efficient and friendly interface.
